Three experiments examined the reciprocity of evaluative effects following CS-US pairing. In all three experiments, CS evaluations were assimilated to the valence of the US they were paired with (i.e., an evaluative conditioning effect), whereas US evaluations became less extreme (i.e., a US devaluation effect). Of importance, however, US devaluation proved to be independent of CS-US pairing. Experiment 1 replicated previous evidence for US devaluation: USs were less intensely evaluated after a conditioning procedure as compared to their normative ratings. Experiment 2 controlled for the effect of CS-US pairing: A US devaluation effect of similar magnitude was observed for USs paired with the CSs or presented alone during the conditioning procedure. Experiment 3 indicated that US habituation drives US devaluation: USs presented and evaluated only once were less devalued than USs paired with CSs or USs presented alone during the conditioning procedure, with the latter two US types not differing from each other. Together, these findings suggest that US devaluation is driven by US habituation rather than by a CS-to-US influence in an associative learning procedure. The theoretical implications of these findings for associative and propositional accounts of evaluative learning are discussed.
In general, people approach what they like and avoid what they dislike, and this is true in many aspects of their life, including social interactions (
That the US elicits an evaluation prior to its pairing with the CS does not imply that its evaluation is not to be altered following that pairing. The US, too, may be conditioned (i.e., it may also be a conditioned stimulus). Consistent with this view,
The US devaluation effect has important practical and theoretical implications. One practical implication is clinical in nature: if people acquire a strong dislike for certain categories of stimuli (e.g., spiders), is there any advantage in actively pairing these with neutral stimuli, rather than presenting them in isolation?
At the theoretical level, the US devaluation question is relevant to associative versus propositional attitude learning views. As compared to propositions, associations are “unqualified” in nature. As such, they are assumed to contain no structural content qualifying the CS-US relation (
In order to address these questions, it is important to proceed first to a closer examination of the processes that may underlie the US devaluation effects. In particular, two processes may account for US devaluation that are independent of any effect of CS-US pairing. First, US devaluation may reflect a regression to the mean effect. If a variable is extreme upon its first measurement, it will tend to regress to the average on a subsequent measurement (
The present experiments aimed to (1) test the robustness of the US devaluation effect reported by
Experiment 1 conceptually replicated
Experiment 1 allowed testing of the presence of the US devaluation effect reported by
136 participants (
The procedure involved three phases. In the first - conditioning - phase, eight CS-US pairs were displayed on the screen. Four CSs were presented with positive USs and four CSs were presented with negative USs. The specific CS-US valence pairings were created randomly for each participant. Each pair was presented eight times, in a random order, for 1000ms. The Inter-Trial-Interval (ITI) lasted 500ms. CSs and USs were borrowed from
For each participant, we computed both for the CSs and for the USs the average difference in ratings between post-conditioning ratings and normative ratings. Normative ratings for the CSs were collected in a pre-test (see
Changes in CS evaluation were submitted to a repeated-measures ANOVA with US valence (positive vs. negative) as a within-participant factor. The main effect of US valence was significant,
Changes in US evaluation were submitted to a repeated-measures ANOVA with US valence (positive vs. negative) as a within-participant factor. The main effect of US valence was significant, indicating that US evaluations became more neutral after the pairing (i.e., a US devaluation effect),
Change in evaluation for positive and negative CSs and USs. Dots represent individual responses. Filled squares represent observed means. Error bars represent standard errors of means. CSs- = negatively conditioned stimuli; CSs+ = positively conditioned stimuli; USs- = negative unconditioned stimuli; USs+ = positive unconditioned stimuli.
In Experiment 1, the valence of the CSs was more extreme after than before the EC procedure (i.e., an evaluative conditioning effect emerged). Conversely, USs elicited less extreme evaluations after than before CS-US pairing (i.e., the US devaluation effect was replicated). Because of the use of normative ratings as a baseline, no regression to the mean effect could contribute to US devaluation. Instead, US devaluation may be due to the CS-US pairing or to the mere presentation of the USs. Of note, US devaluation was significant only for the positive USs. One should therefore remain cautious about the interpretation of this result as it may be due to the reliance on different samples for normative and experimental ratings. Collecting evaluations before and after CS-US pairings within the same sample of participants allows this interpretational issue to be overcome. This is the procedure used in Experiment 2.
The role of the CS-US pairing was examined in Experiment 2 by including control US-alone stimuli in the conditioning procedure. As a further asset, Experiment 2 relied on idiosyncratic rather than normative US evaluations. In this case, participants served as their own control, although USs were still selected on the basis of prior normative ratings (therefore precluding regression to the mean effects). We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the experiment. The raw data from the experiment as well as the R script created to manage, analyze, and visually represent the data are available on Open Science Framework (OSF link:
We collected data on 163 participants (
Participants first rated eight CSs (
We computed the difference between post- and pre-ratings for the CSs and USs at the stimulus level for each participant. A positive value reflects more positive evaluations at post-test. We then averaged evaluative change scores by participants as a function of US valence and pairing type. As in Experiment 1, we used frequentist and Bayesian repeated-measures ANOVA to analyze the data.
Changes in CS evaluation were submitted to a repeated-measures ANOVA with US valence (positive vs. negative) as a within-participant factor. This analysis revealed a main effect of US valence,
Changes in US evaluative ratings were submitted to a repeated-measures ANOVA with US valence (positive vs. negative) and pairing type (alone vs. paired) as within-participant factors. We observed a main effect of US valence,
Change in evaluations of the USs as a function of US valence and pairing type. Dots represent individual responses. Filled squares represent observed means. Error bars represent standard errors of means. USs- = negative unconditioned stimuli; USs+ = positive unconditioned stimuli.
Experiment 2 replicated both an EC effect and a US devaluation effect. We found no evidence, however, that CS-US pairing effect contributed to US devaluation: US presented alone and US paired with CSs decreased in valence to a similar extent. The Bayesian analysis suggests that the null effect of pairing is over five times more likely than an effect of pairing in the devaluation effect. Once again, the US devaluation was observed only for the positive USs. This time, this cannot be due to a difference in the samples investigated (i.e., normative versus experimental). Instead, this imbalance points to an asymmetric affective habituation effect, with larger habituation to positively than negatively charged stimuli.
Experiment 3 was designed to further investigate the role of CS-US pairing in US devaluation. In addition, this experiment provides a formal test for the contribution of US habituation. This was achieved by comparing evaluations for USs that were presented and rated only once versus USs that were either paired with CSs or presented alone in the conditioning procedure. If US habituation plays a role, ratings should be less extreme for US-alone and US-paired stimuli than for USs presented and rated only once. We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the experiment. The raw data from the experiment as well as the R script created to manage, analyze and visually represent the data, are available on Open Science Framework (OSF link:
We collected data on 82 participants (
A total of 24 USs were used (see
We averaged CS and US evaluative ratings within each US valence and pairing type for each participant. Similar to Experiment 1, we then computed the average difference in ratings between post-conditioning ratings and normative ratings as a function of US valence and the presentation condition. As for the previous experiments, we used frequentist and Bayesian repeated-measures ANOVA to analyze the data.
Changes in CS evaluation were submitted to a repeated-measures ANOVA with US valence (positive vs. negative) as a within-participant factor. This analysis revealed a main effect of US valence,
Changes in US evaluation were submitted to a repeated-measures ANOVA with US valence (positive vs. negative) and pairing type (US-never vs. US-alone vs. US-paired) as within-participant factors. This analysis revealed no main effect of US valence,
This analysis further revealed a US valence × pairing type interaction,
Change in evaluations of the USs as a function of US valence and pairing type. Dots represent individual responses. Filled squares represent observed means. Error bars represent standard errors of means. USs- = negative unconditioned stimuli; USs+ = positive unconditioned stimuli.
Specifically, the difference in the mean change of US evaluation between positive and negative pairings was observed for both US-alone trials,
Looking at habituation effects within each US valence, ratings between learning conditions differed for both positive USs,
In Experiment 3, USs presented alone or with a CS during the conditioning procedure were evaluated less intensely (and equally so) than USs never presented before. Hence, the present study again finds no evidence for the contribution of CS-US pairing in the US devaluation effect. In addition, it provides evidence for the contribution of the mere US presentation to this effect. Finally, consistent with Experiment 1 and 2, a US devaluation effect was found for positive USs only.
In this research, we examined the reciprocity of evaluative changes following CS-US pairings. An evaluative conditioning effect emerged systematically, which confirms the influence of the US in the evaluation of the CS after pairing these together. Consistent with
Across studies, habituation was systematically observed for positive USs only. As for the negative USs, there was a trend for habituation in Experiments 1 and 3, but it was absent in Experiment 2. The asymmetry in valence of affective habituation echoes the asymmetry observed in the processing of positive versus negative information. Overall, two types of such affective asymmetries have been documented, both of which may account for the present findings. First, a “good is weaker than bad” account would suggest that negative information has a stronger psychological impact than positive information. Typically, negative information draws more attention (
The present findings suggest that evaluative changes elicited by CS-US pairings are non-reciprocal, in the sense that pairing the US with a CS has no demonstrated influence on the US evaluation over and above that elicited by US habituation. The one-sided evaluative effect of the CS-US association seems problematic for dual-learning models of attitudes that rely on associative assumptions. Specifically, it is unclear why unqualified associations between stimuli should show non-reciprocal effects (see also below). One possibility is that CSs, being neutral, elicit no response, the latter of which would be needed for establishing evaluative effects. Whether an absence of affective response should be equated to the presence of a neutral response is an interesting question. To the extent that non-valenced stimuli are identified by neutral evaluative ratings (i.e., responses), the answer to this question is affirmative: neutral stimuli, clearly, do elicit (neutral) evaluative responses. It is therefore important to specify the scope of observable responses elicited by neutral stimuli that would or would not be relevant to associative or dual-learning models of attitude. The latter specification would help to further constrain these models.
Alternatively, CS-to-US effects may have been too small to allow their detection in Experiment 2 and 3. Although it is possible that the present experiments did not reach sufficient power to detect such small effects, we systematically observed substantial evidence in a Bayesian approach for the null-hypothesis that CS-US pairings have no effect on US evaluation over and above the mere US presentation (see, for instance,
If we cautiously assume that effects of CS-US pairings are non-reciprocal, then this would suggest that the evaluative effect of CS-US linkage is sensitive to a directional qualifier, one that goes from the US to the CS only, at least by default. This qualified (in this case, non-reciprocal) CS-US linkage is compatible with propositional accounts of attitude learning (e.g.,
By comparison, the absence of a reciprocal effect seems less consistent with an associative learning account. This is because associative learning typically refers to a low-level process that automatically registers co-occurrences between stimuli independent of their validity and relational meaning (for a recent discussion, see
The data from all three experiments as well as the R scripts created to manage, analyze, and represent the data visually are available on Open Science Framework (see the
The raw data and analysis scripts created to manage, analyze, and represent the data visually are available online via the project hub on the Open Science Framework:
USs - | USs + |
---|---|
6312, 2750, 6360, 6561, | 4608, 4700, 8200, 8460, |
2141, 2900, 6315, 6510, | 2550, 4603, 4641, 8120, |
6210, 6312, 6550, 6571 | 1603, 2501, 8162, 7325 |
The authors have no funding to report.
The authors have declared that no competing interests exist.
The authors have no support to report.